Grok Voice Mode enters internal testing as it gears up to rival ChatGPT and Gemini

xAI is preparing the Voice Mode for launch, and it is already functional in the latest iOS app, though still hidden from users. When enabled, it works smoothly and matches the interface showcased during the official Grok 3 announcement. This design differs from previously leaked versions, offering a more polished and refined user experience.

Currently, there is only one available voice called Sal, a male voice. However, there is a dropdown menu within the interface, indicating that more voice options might be added in the future. The user interface includes three main buttons at the bottom:

• A Close button to exit the voice mode
• A Mute button to silence the conversation
• A Share button, expected to function similarly to ChatGPT’s clip-sharing feature, allowing users to share recordings of their voice conversations

In addition to these features, there are three more key elements:

• A toggle gives users the option to turn Internet access on or off during the voice session
• A Conversation Transcription option, which displays real-time transcriptions of the voice chat
• A Settings Menu, where users can access a custom prompt feature, potentially designed specifically for voice mode interactions

This is particularly useful, as voice and text conversations often require different instruction styles. However, it remains unclear if the transcription and custom prompt features will be available from the start.

In terms of functionality, the voice mode can sing songs, express emotions, and handle dynamic conversations. Despite these capabilities, it is expected to be more restricted than ChatGPT’s Advanced Voice Mode or Gemini Live. Importantly, this is a native voice mode powered directly by the Grok 3 model, not a simple text-to-speech system—this was confirmed during the Q&A session following the official demo.

According to Elon Musk, the feature is expected to launch on Grok within one to two weeks, as the xAI team continues to fine-tune it. While it appears ready, it will likely enter an internal testing phase with friends and family before a public rollout. Given its current state, an official announcement seems imminent.